Scattering and absorption in soft X-ray selected AGN: an optical polarization survey

Transcription

1 Astron. Astrophys. 333, (1998) Scattering and absorption in soft X-ray selected AGN: an optical polarization survey D. Grupe 1,2,, Beverley J. Wills 1, D. Wills 1, and K. Beuermann 2 1 Department of Astronomy & McDonald Observatory, University of Texas at Austin, RLM , Austin, TX 78712, USA 2 Universitäts-Sternwarte, Geismarlandstrasse 11, D Göttingen, Germany ASTRONOMY AND ASTROPHYSICS Received 19 August 1997 / Accepted 22 December 1997 Abstract. We have surveyed the optical linear polarization of a completely identified sample of 43 bright soft-x-ray-selected ROSAT AGN. Most (40) of these AGN show low polarization ( <1%), and no clear optical reddening. This supports the suggestion from rapid X-ray variability, disk-like spectral energy distributions, and lack of cold X-ray absorption, that we are viewing a bare AGN disk. IRAS F and IRAS show high polarization increasing to the UV clear evidence for scattering. As well as steep, soft- X-ray spectra, they show optical reddening and rapid X-ray variability, but almost no cold X-ray absorption a combination that suggests dusty ionized gas along the line-of-sight. Brandt et al. suggested and found these warm absorbers for IRAS IRAS F is a new candidate. Combining our data with the optical and X-ray spectra of the high polarization narrow-line Seyfert 1 nuclei (NLSy1s) investigated by Goodrich reveals strong correlations among optical reddening indicators (α opt and Hα/Hβ), [O III]/Hβ b, and cold intrinsic X-ray absorption N H. Optical reddening underpredicts the cold X-ray absorption, suggesting dusty warm absorbers in all the highly polarized AGN. The existence of these scattering-polarized and reddened NLSy1s suggests an orientation Unified Scheme within the class of NLSy1s, analogous to that linking Seyfert 1 and Seyfert 2 nuclei. For some highly polarized and optically selected AGN we present new analysis of archival X-ray data, and for the highly polarized AGN new optical spectroscopy is presented in an appendix. Key words: polarization galaxies: active galaxies: nuclei galaxies: Seyfert 1. Introduction A popular idea among astronomers is that all AGN are powered by the same kind of engine: accretion of matter through an accretion disk onto a super-massive black hole. This idea Send offprint requests to: D. Grupe (dgrupe@xray.mpe.mpg.de) Current address: MPE, Giessenbachstr., D Garching, Germany receives strong support from high, Keplerian velocities of gas in the centers of active galaxies, and alignment of the axes of these gaseous disks with powerful radio jets. Understanding of the basic fueling mechanism remains elusive, but there may be new clues, related to the classic energy-budget problem and the so-called First Principal Component, both linking the ionizing continuum and the emission-line regions: (a) The energy budget problem is that there appears to be insufficient ionizing continuum photons to power the broad emission lines, especially the strong, blended Fe II lines (e.g. Netzer 1985; Collin-Souffrin et al. 1988). Solutions may be that the Fe II arises from a different region, excited not by photoionization but by mechanical heating of the gas (e.g. Norman & Miley 1984; Joly 1991), or that the continuum incident on the broad emission-line gas is different from that observed. (b) In AGN s UV-optical spectra there are several clear, related trends: increasing broad-line Fe II emission corresponds to decreasing width (FWHM) of Hβ from the broad-line region (BLR), and to decreasing line emission from lower velocity, lower density gas [O III] λ5007 emission from the narrow line region (NLR), and UV line emission from lower velocity gas of the BLR (Boroson & Green 1992; Laor et al. 1997; Lawrence et al. 1997; Brotherton et al. 1997, in preparation). ROSAT, with its high sensitivity to soft X-rays, has played an important role in showing that these trends correspond to increasing steepness of the soft X-ray spectra (Boller et al. 1996, Laor et al. 1997, Grupe et al. 1998, in preparation). Correlation analyses show that the dominant spectrum-to-spectrum variation in low and high redshift AGN samples can be reduced to a linear combination of these emission line and continuum parameters the First Principal Component. This First Principal Component points to a single, as yet mysterious, underlying mechanism relating AGN s ionizing continuum ( kev), the optical Fe II emission, and emission from the BLR and NLR. While rarely discovered in optical surveys, narrow-line Seyfert 1 nuclei (NLSy1s) are found in abundance in soft-xray surveys. These AGN, having broad Hβ emission lines with FWHM < 2000 km s 1, and steep, strong, soft-x-ray spectra, lie at the rich-fe II extreme of the First Principal Component, and so could give us insight into central engine mechanisms. One hypothesis proposed to explain their strong, often timevariable, soft X-ray emission, their steep X-ray spectra and nar-

2 828 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN row BLR Hβ emission lines, is that NLSy1s represent unobscured, low-mass AGN (Boller et al. 1996; Grupe 1996; Grupe et al. 1998, 1998, in preparation). Lower mass means lower gas velocities, producing narrower Hβ emission from the BLR. The strong, soft X-ray-emission arises from an unobscured view of the inner regions of a smaller, and therefore hotter, accretion disk. This direct, rather than scattered-light, view of the tiny emission region allows very rapid variability. Further support for this hypothesis is an observed anticorrelation between optical and soft-x-ray spectral indices, and, in the most luminous AGN, an optical continuum slope equal to that expected from optically thick, thermal, disk emission. On this hypothesis, NLSy1s should be unpolarized. Another hypothesis suggests that NLSy1s are Seyfert 1s viewed pole-on. Kinematic projection of a disk-like BLR results in a narrower radial velocity distribution, and steep X-ray spectra result from a view of the funnel of a geometrically thick accretion disk (Madau 1988). Perhaps some fraction of these NLSy1s are like the archetype of highly polarized QSOs, IRAS (Wills et al. 1992b, Brandt et al. 1996). The soft X-ray spectrum of this AGN is strong, variable, and steep, with no signs of absorption by neutral gas. In the optical, it shows typical NLSy1 characteristics narrow Hβ from the BLR, very strong Fe II, and very weak [O III] λ5007. In this AGN there are two views to the central continuum source and BLR one a direct view that is partially obscured by dusty hot gas in the torus, and the other a less obscured, but indirect, view in which the spectrum is scattering-polarized by electrons or dust in the opening cone of the torus. The unusual combination of UV-optical reddening and lack of cold X-ray absorption was presented by Brandt et al. (1996) as evidence for dusty, ionized gas near the nucleus. Discovery of more such objects would allow further optical- UV-X-ray investigation of this newly discovered constituent of AGN gas. The above pictures for NLSy1s are not mutually exclusive. In axisymmetric Unified Schemes, a low-mass or thickdisk central engine may be surrounded by a dusty nuclear torus. A low-inclination polar view may reveal an unobscured central engine, while a higher inclination line-of-sight may graze the dusty nuclear torus, passing through ionized gas (e.g. IRAS ). As dust further dims the direct view of the nucleus, the scattered, polarized AGN spectrum becomes more prominent. At even higher inclinations, cold dusty gas blocks the direct view completely, but a few percent of the NLSy1 spectrum could be scattered towards the observer. Even in a high inclination view, low density NLR gas, ionized on kpc scales, may reveal the presence of a buried NLSy1 in total light. The well-known Seyfert 1 nucleus buried within the prototype Seyfert 2 galaxy NGC 1068, the Rosetta Stone of Unified Schemes (Antonucci & Miller 1985), may be an excellent highinclination example with the parsec-scale AGN seen only in scattered light a faint, polarized spectrum with UV-optical continuum, strong Fe II blends, and BLR Hβ intrinsic FWHM 2900 km s 1 (Miller et al. 1991), together with a very steep, unabsorbed soft X-ray spectrum (α x 2.4), showing no variability on several-year time-scales (Elvis & Lawrence 1988; Marshall et al. 1993; Smith et al. 1993). In this paper we have used broad-band polarimetry and spectroscopy of a ROSAT sample of soft-x-ray-selected AGN to relate scattering and absorption in the optical and X-rays, in order to investigate anisotropic emission, kinematics and dusty gas distribution in these special AGN. In Sect. 2 we describe the sample, as well as observations and data reduction for the optical polarimetry, ROSAT X-ray spectroscopy, and optical spectroscopy. Sect. 3 presents the results, including some correlations. Here we provide more detailed information for the highly polarized sources, and collect together comparable X-ray and optical data for the AGN of Goodrich s (1989) spectropolarimetry study of optically selected NLSy1s. Sect. 4 discusses the absorption and scattering properties of the sample as a whole, and for individual sources, including a comparison with the optically selected NLSy1s. A summary is given in Sect. 5. Appendix A shows our spectra for all the highly polarized sources discussed in this paper. 2. Observations and data 2.1. The soft X-ray sample The sample of bright, soft-x-ray AGN (Grupe et al. 1998) was selected from the ROSAT All-Sky Survey (RASS, Voges et al. 1993, 1997). This deep survey, using the Position Sensitive Proportional Counter (PSPC, Pfeffermann et al. 1986), was the first to extend to soft-x-ray photon energies 0.1 kev. Our X-ray sample is completely identified, comprising all 95 AGN with RASS PSPC count rates >0.5 cts s 1, hardness ratio HR1 <0.0, and b > 20. HR1 is defined as the ratio of the difference to the sum of hard and soft counts of the source, where the hard and soft bands are defined between kev and kev. This HR1 limit corresponds to a steep soft X-ray spectrum with α x > 1.5 for N H cm 2. The bright, soft X-ray selection results in a sample with very little or no neutral hydrogen absorption. The X-ray and optical properties of the sample were presented and discussed by Grupe (1996) and Grupe et al. (1998, and 1998 in preparation). About half the sample are NLSy1s. The median FWHM for broad Hβ is 2250 km s 1, with a 90-percentile range from 1300 km s 1 to 4200 km s 1. We restrict the present polarization survey and analysis to the 43 northern sources (δ >0 ) Polarimetry observations For the soft X-ray sample, linear polarization was measured using a broad-band polarimeter (Breger 1979) on the Struve 2.1-m telescope at McDonald Observatory. The detector was a cooled ( 20C) Hamamatsu R phototube with extended red sensitivity [GaAs(Cs) photocathode] operated in photon-counting mode. A new interface and data acquisition computer resulted in improved noise characteristics with uncertainties consistent with photon statistics. Otherwise, the filters, observational technique (a few 200-second integrations onsource, interleaved with background integrations), calibrations,

3 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN 829 and reductions were as described by Wills et al. (1992a). All measurements were made with an aperture of projected diameter 7.4. We were interested in searching for scattering polarization, so we aimed for 1σ measurement uncertainties <0.4%, because interstellar polarization mechanisms may become important at that level, and are then difficult to distinguish from scattering, based on degree of polarization and wavelength-dependence. In some cases where the measurements indicated real polarization we estimated the contribution from Galactic interstellar polarization by measuring the polarization of stars nearby in projection on the sky. In order to probe as long a path length as possible through our galaxy we chose (where possible) bright, blue stars, preferably known to be of early spectral type. Of the 43 northern AGN, 26 were observed by us and data for 17 were taken from the literature X-ray data The RASS data have been used for the statistical investigation of the soft-x-ray sample, except for well-known AGN. Details of the reductions are given by Grupe (1996) and Grupe et al. (1998). For the 19 well-known sources we retrieved pointed observation data from the ROSAT public archive at MPE Garching. The January 1996 version of the EXSAS package (Zimmermann et al. 1994) was used for data reduction and analysis. The X-ray spectral index α x (F ν ν αx, the photon index Γ=1+α x ) was determined in the energy range kev using single power-law fits, assuming neutral absorption given by Milky Way HI 21cm column densities N H,gal (Dickey & Lockman 1990) with abundances given by Morrison & McCammon (1983). Uncertainties in α x are typically 0.1 to 0.5, depending on X-ray brightness. Grupe et al. (1998) also made fits to the X-ray spectra with unconstrained neutral hydrogen column-density N H. Although the latter was rarely significantly different from Galactic values, we measured the difference N H = N H N H,gal as a potential indicator of intrinsic neutral hydrogen absorption. For those objects for which no X-ray spectra were readily available, we used the hardness ratios HR1 and HR2 given in the bright source catalogue (Voges et al. 1997). HR2 is like HR1, but with hard and soft bands defined between kev and kev. The method is to simulate count-rate spectra for a grid of α x and N H values, using the known instrumental response. From these spectra we calculate HR1 and HR2 for each combination of α x and N H. Thus we derive the unique combination α x and N H that reproduce the observed HR1 and HR2. Uncertainties in N H are similar to those in N H determined from the fits with N H unconstrained. The standard deviations for individual values are about cm 2 and cm 2 as determined from the median and mean of the distribution of standard deviations given from the fitting. For the highly polarized AGN for which ROSAT pointed observations were available, the data were analyzed in the same way as for the RASS data Optical spectroscopy In general, we used the medium-resolution optical spectra (FWHM 5Å) from Grupe (1996) and Grupe et al. (1998, in preparation), obtained in March The highly polarized objects were re-observed, including Hβ and Hα, in March 1997 with the 2.1m telescope at McDonald Observatory, but with resolution FWHM 7Å. Spectra for the PG quasars (resolution 3.6Å FWHM) are from observations with the 2.7-m telescope at McDonald Observatory in February A spectrum for RE J is from Puchnarewicz et al. (1995a), and for CSO 150, from Bade et al. (1995). Except for QSO , and the higher redshift AGN RX J and RX J , we were able to find spectra in the literature: MS (Stephens 1989), EXO (Bedford et al. 1988), UM 472 (Salzer et al. 1989), and NGC 4593 (Dietrich et al. 1994). The emission line measurements and their uncertainties are presented and described in detail by Grupe et al. 1998, in preparation). Some representative 1σ errors are given in later tables. The blended optical Fe II emission was measured and subtracted using a I Zw 1 template like that used by Boroson & Green (1992). All other measurements were made from these Fe II-subtracted spectra. To determine the rest-frame equivalent width of Fe II we measured the flux in the template between rest wavelengths 4250Å and 5880Å, after scaling it to match the object spectra, and then divided by the continuum flux density at 5050Å. In the following, Hβ refers to the whole Hβ line. Hβ b refers to the broad component of Hβ, obtained after subtraction of the narrow component by using the [O III]λ5007 velocity profile as a template. This method subtracts as much of the template as possible while still retaining a sensible shape to the peak of Hβ b. For sources where the narrow-line emission dominates we found an [O III]/Hβ narrow intensity ratio of about 10, which is not significantly different from the observed median in Seyfert 1s and Seyfert 1.5s (Koski 1978; Cohen 1983), providing support for our method. To determine the Hα/Hβ intensity ratio, we used the total flux in the lines, because the broad and narrow components are difficult to separate for the Hα blend. In general, to subtract the fluxes of the [N II] λλ6548,6584 lines from the Hα blend, it was quite adequate to subtract 35% of the [O III]λ5007 line flux from it (Ferland & Osterbrock 1986). For the highly polarized sources, as we see later, reddening of the NLR emission could be important. So for these we used velocity templates derived from [O III] and total Hβ lines to disentangle the Hα [N II] blend. Optical continuum spectral indices α opt were measured between rest-frame wavelengths of 4400Å and 7000Å The optically selected sample For comparison with our soft-x-ray-selected sample, we have compiled polarization, spectroscopic and X-ray data for the optically selected sample of NLSy1s that Goodrich (1989) investigated by spectropolarimetry. While the selection of that sample

4 830 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN Fig. 1. Distribution of the polarization of the 43 northern soft-x-ray AGN. is based on a mixture of spectroscopic criteria (narrow Hβ emission, strong Fe II emission, [O III]λ5007/Hβ <3, and presence of high-ionization lines) it is unbiased with respect to soft-xray properties (α x or N H ). Measurements were made, as far as possible, in the same way as for our sample. Emission-line measurements were derived from literature data, except for the more highly polarized sources for which we obtained new spectroscopic data including both the Hβ and Hα regions (Appendix A). 3. Results 3.1. The ROSAT soft X-ray AGN sample The polarimetry data are given in Table 1, where we list the object s RX J name derived from the RASS J2000 coordinates, other name, the filter used, degree of polarization p%, and the polarization position angle (E vector, measured on the sky plane in degrees from north through east), and a literature reference where relevant. Errors in p% are 1σ, based on photon-counting statistics. The corresponding position-angle error may be estimated from the fractional error in p%. No corrections have been applied for positive bias in p%. Neither have we made any attempt to correct for unpolarized starlight from the host galaxy. Table 2 gives measurements we made for interstellar polarization. In no case was this large enough to justify a correction to our AGN polarizations. Fig. 1 displays the distribution of the polarization. Most of the soft-x-ray AGN are of low polarization ( <1%). Three of the 43 objects show p >1%. While we have chosen 1% as a practical cut-off for investigating scattered-light polarization, we note that two other AGN, with detailed results from the literature, show low but significant polarization, above interstellar levels. One is the well-known NLSy1 NGC The other is the broad-lined NGC 5548, with HR1 0. Because one motivation of this study was to investigate Unified Schemes for NLSy1s, we ask whether there are systematic differences between high and low polarization AGN. The FWHM(Hβ b ) of the highly polarized AGN are not extremely Fig. 2. The optical spectral index α opt vs. total Hα/Hβ. Data for our northern and southern X-ray-selected samples are shown as circles, with IC 3599 shown as an open square. See Table 4 for representative 1σ uncertainties. The majority, thought to be unreddened, are shown as open circles. The highly polarized AGN (p >1%) are shown as filled symbols circles for the soft-x-ray AGN, triangles for the opticallyselected sample. CBS 126, with p 1%, is shown as a heavy circle. The tick-marked line in the plot displays values of E B V for equal line and continuum reddening. Polarization data are not available for the majority of southern sources. narrow but typical for the soft X-ray sample. Also, α x 1.7 for the highly polarized AGN, so their soft X-ray spectra tend to be flat compared with most of our sample; but steep compared with ROSAT spectra of optically-selected Seyfert 1 nuclei, for which α x (Grupe et al. 1998). Reddening of the broad lines and continuum is another test for whether we have a direct view to a bare nucleus, with the narrow Hβ b and steep α x being intrinsic properties, or whether soft-x-ray-selected AGN with narrow Hβ b may be viewed through dusty gas near the nucleus, as in orientation Unified Schemes. In Fig. 2 we compare α opt and the Balmer decrement, as indicators of reddening. The low polarization AGN (most of the open circles) have small Hα/Hβ ratios, and flat optical spectra (small α opt ), and show no relation between Hα/Hβ and α opt outside the expected relation resulting from systematic calibration uncertainties, so reddening is unlikely to be important for them. While the low polarization sources do not show significant reddening, the highly polarized sources do. All the polarized AGN (filled symbols) have Balmer decrements >6 (except CBS 126, which we discuss later). The one low polarization source with Balmer decrement >5.8 is IC 3599, which is thought to be a Seyfert 2, or more likely, a starburst nucleus (Sect. 4.4) and therefore different from the rest of our sample. Another test is to look for intrinsic absorption N H in the X-ray spectra. As remarked before, RASS data did not show significant absorption in soft X-rays for any AGN, except

5 Table 1. Polarimetry of Soft X-ray-selected AGN. D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN 831 Object Other Polarization Filter RX J Name p% Angle ( ) Source of Polarimetry data MS none 0.75± McD2.1 March Mkn 110 none 0.21±0.13 Berriman et al. (1990) PG none 0.25±0.22 Berriman et al. (1990) PG none 0.77±0.22 Berriman et al. (1990) none 0.40± McD2.1 April CBS 126 none 1.26± McD2.1 April 1996 CuSO4 1.19± McD2.1 April none 0.03± McD2.1 April none 0.42± McD2.1 April Mkn Å 0.08±0.28 Berriman (1989) Å 0.29±0.28 Martin et al. (1983) Mkn 142 none 0.11±0.27 Berriman et al. (1990) RE J none 0.15± McD2.1 March Å 0.36± Breeveld & Puchnarewicz (1997) Å 0.40± Breeveld & Puchnarewicz (1997) none 0.31± McD2.1 April EXO none 0.44± McD2.1 April PKS none 0.63± Wills et al. (1992a) none 0.40± McD2.1 April PG none 0.37±0.24 Berriman et al. (1990) Ton 1388 none 0.23±0.11 Berriman et al. (1990) Mkn Å 0.32±0.13 Berriman (1989), Martin et al. (1983) none 0.39±0.17 Berriman et al. (1990) EXO none 0.29± McD2.1 March QSO none 0.33±0.28 McD2.1 March Z none 0.48± McD2.1 April CSO 109 none 0.03± McD2.1 April NGC Å 0.66±0.10 Berriman (1989), Martin et al. (1983) UM 472 none 0.62± McD2.1 March PG none 0.07±0.09 Berriman et al. (1990) none 0.86± McD2.1 April CSO 150 none 0.20± McD2.1 March IC 3599 none 0.24± McD2.1 April NGC Å 0.49±0.16 Berriman (1989), Martin et al. (1983) IRAS F none 3.77± McD2.1 April 1996 U 7.00± McD2.1 April 1996 CuSO4 5.20± McD2.1 April 1996 B 4.99± McD2.1 April 1996 V 3.98± McD2.1 April 1996 R 3.50± McD2.1 April 1996 I 2.16± McD2.1 April none 0.20± McD2.1 April none 0.30± McD2.1 February PG none 0.81±0.22 Berriman et al. (1990) IRAS none 4.63± Wills et al. (1992b) none 0.53± McD2.1 April PG none 0.29±0.14 Berriman et al. (1990) none 0.20± McD2.1 April NGC Å 0.71±0.10 Berriman (1989), Martin et al. (1983) Mkn 684 none 0.18± Goodrich (1989) Mkn Å 0.43±0.15 Berriman (1989) Å 0.46±0.15 Martin et al. (1983) none 0.26±0.17 Berriman et al. (1990) none 0.35± McD2.1 February PG none 0.59±0.19 Berriman et al. (1990) none 0.39± McD2.1 April 1996

6 832 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN Table 2. Measurements for interstellar polarization. Right ascension and declination offsets α and δ in arcminutes relative to the AGN coordinates AGN Star Polarization α δ Filter Name Name % Angle CBS 126 SAO none 0.10± Star nf none 0.10± RX J SAO none 0.11± IRAS F SAO none 0.04± SAO V 0.03± R 0.06± I 0.09± Fig. 3. Line ratio [O III]/Hβ b vs. equivalent width of the optical Fe II blend. See Table 4 for representative 1σ uncertainties. The symbols are as defined in Fig. 2. The dotted vertical lines show the de-reddening paths, based on E(B-V) values derived from Fig. 2. IC However, for some of the highly polarized AGN the higher signal-to-noise pointed observations do show probable absorption (Sect. 3.3) We can also test how the highly polarized sources might be related to the First Principal Component. In Fig. 3 we show the line ratio [O III]/Hβ b vs. equivalent width of the optical Fe II blend, EW(Fe II). The low polarization sources of the present sample show the well known inverse [O III] Fe II trend seen for radio-quiet AGN as a whole (Boroson & Green 1992, Grupe et al. 1998, in preparation). Our high polarization sources appear to lie at larger [O III]/Hβ b or larger EW(Fe II) than the low polarization sources (see Sect. 4) Comparison with an optically selected NLSy1 sample The data for the optically selected NLSy1 sample are presented in Table 3 where we list the object s common and RX J name, percentage polarization and angle, X-ray count rates and hardness ratio, α x, N H, FWHM(Hβ b ) and [O III]/Hβ b ratio. The polarization data are mean values over the whole observed wavelength range (Goodrich 1989). Because soft-x-ray absorption is more probable in this non-soft-x-ray-selected sample we give the α x determined from an X-ray spectral fit with N H unconstrained. Sources of data are given below the table. As expected, the optical selection criteria for this sample result in narrower Hβ b and a wider range of HR1 compared with the X-ray-selected sample. Only half of the optically selected sample have steep α x > 1.5. Out of 17 objects, several are polarized at a low level, but three show continuum p>1%, even >2%: Mkn 1239, Mkn 766, and IRAS As for the highly polarized sources of the soft-x-ray sample, these three have FHWM(Hβ b ) typical of the unpolarized or low-polarization AGN, but they are among the most reddened and have among the highest [O III]/Hβ b ratios (Figs. 2 & 3). Mkn 1239 and IRAS have the largest HR1 of the sample, apparently the result of soft-x-ray absorption. Mkn 783, while not highly polarized, has a high HR1 and soft X-ray absorption, and the largest [O III]/Hβ b The polarized AGN Two of the three highly polarized AGN of our sample show high and wavelength-dependent polarization IRAS F (Was 61, Wasilewski 1983) and IRAS (Fig. 4). The IRAS polarization has been the subject of intensive investigation as a result of its membership in a warm IRAS sample (Sect. 1). While it provided motivation for the present survey, its inclusion in our statistical discussion is justified because it was discovered by completely independent techniques and was not appreciated to be a member of our sample until much later. But IRAS F is new. The strong increase of polarization towards shorter wavelengths is a clear indication of the dilution of a scattered light AGN spectrum by a redder unpolarized component either reddened AGN light or starlight. The other highly polarized source, CBS 126, appears to be in a different category. The polarization is small and not strongly wavelength dependent if at all. Unlike the other polarized AGN, the broad lines and continuum are not significantly reddened, and the larger FWHM(Hβ b ) is more typical of normal Seyfert 1

7 Table 3. Properties of Goodrich s optically selected NLSy1 sample D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN 833 Common Polarization CR i i N H FWHM(Hβ b ) RX J i ii Name % Angle cts s 1 HR1 i i [O III] α x cm 2 km s 1 Hβ b Mkn ± (a) +0.0 (a) 685 (a) 0.40 (a) Mkn ± (b) +0.3 (b) 1350 (b) 1.73 (a) Mkn ± (b) +1.1 (b) 1500 (c) 0.22 (a) Mkn ± (b) +4.4 (b) 1050 (d) 2.57 (d) PG ± (a) 0.10 (a) Mkn ± (b) +0.6 (b) 670 (e) 0.35 (a) Mkn ± (b) +0.4 (b) 1360 (d) 3.82 (d) NGC ± (a) +0.0 (a) 1470 (a) 1.60 (a) Mkn ± (a) +2.8 (a) 1900 (e) 9.42 (a) Mkn ± (c) +0.4 (c) 1690 (f) 0.16 (f) IRAS ± (d) (d) 2250 (d) 1.16 (d) Mkn ± (b) 1.4 (b) 700 (g) 0.98 (a) Mkn ± (b) +0.6 (b) 1360 (d) 0.35 (d) VII Zw ± (a) 1.2 (a) 1260 (a) 0.53 (a) Mkn ± (b) +0.5 (b) 965 (a) 0.73 (a) Akn ± (b) +1.2 (b) 1000 (b) 1.10 (a) Mkn ± (a) 0.2 (a) 2500 (e) 4.15 (a) ii i X-ray data sources: The count rates CR, and HR1 are from the RASS Bright Source Catalog (Voges et al. 1997). All α x are from single power law fits with N H unconstrained. For those objects for which no entry was found, the RX J coordinates were taken from accurate optical positions. (a) derived from HR1 and HR2 (see Sect. 2.3) (b) Boller et al. (1996) (c) RASS observation (Grupe et al. 1998) (d) ROSAT pointed observation retrieved from the archive ii Optical data sources: (a) Goodrich (1989) (b) Osterbrock & Shuder (1982) (c) Rafanelli (1985) (d) McD2.1 March 1997 (see Appendix A) (e) Osterbrock & Pogge (1985) (f) McD2.1 March 1994 (Grupe et al. 1998, in preparation) (g) Lipari et al. (1993) nuclei. Possibly the polarization arises from transmission of nuclear light through well-aligned grains. For all these reasons, we exclude this source from detailed discussion. The three optically selected NLSy1s with p>2% also show p% increasing to short wavelengths, which, in these AGN, arises from dust-scattered continuum and broad lines (Goodrich 1989). Further details of the X-ray and optical properties of the highly polarized AGN are given in Table 4. The AGN are approximately in order of increasing p% and presumed obscuration. For comparison, in the same table, we include the medians and 90 percentile ranges of all measured quantities for the complete soft-x-ray-selected sample. We give α x and N H based upon the best available X-ray data. For all but CBS 126 and IRAS , the X-ray fits with N H fixed at N H,gal were unacceptable, so for these others we have adopted the α x determined from the power-law fits with N H unconstrained. All scattering-polarized AGN show α opt and Hα/Hβ indicative of reddening by dust (Fig. 2, Table 4), and these two quantities are well correlated. Except for IRAS , the degree of polarization is correlated with reddening, but any such real relation must be complicated by differential reddening of scattered and direct light, different amounts of starlight dilution, and the sensitivity of polarization to details of the non-spherical, projected geometry. Greater optical reddening appears to be well correlated with stronger cold absorption in soft-x-rays as indicated by N H (Table 4). IRAS , with nearly 3 magnitudes of absorption in the optical, is nearly completely absorbed in soft X-rays below 0.5 kev. We should be careful here. The true soft-x-ray absorption may be even greater than indicated by N H if a soft excess is present at the lowest photon energies the kind attributed to a very hot accretion disk (Puchnarewicz et al. 1995b; Grupe et al. 1998). Except for IRAS , the scattering-polarized AGN all have high [O III]/Hβ b ratios (Fig. 3). Alternatively, one could describe the highly polarized sources in this figure as having high EW(Fe II) for a given [O III]/Hβ b ratio (4.2.1). Like the X-ray-selected AGN as a whole, the highly polarized AGN show rapid, large-amplitude, time-variability in soft X-rays. IRAS has shown 50% variations over 6000 s (Brinkmann et al. 1996, Brandt et al. 1997). For IRAS F we find 50% variations over 20 ks. Mkn 766 varied by a factor 2 in 1000 seconds (Leighly et al. 1996). Mkn 1239 varied between the RASS and pointed observations about two years later. The count rate doubled and α x changed from 1.69 during the RASS to 1.94 in the pointed observations, as derived from single power law fits with N H

8 834 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN Table 4. Properties of the highly polarized AGN. Those of our soft-x-ray sample, together with the median and the 90% ranges for the whole soft X-ray sample are to the left. The numbers involved are given in brackets after the median value. The right side of the table gives the values for Goodrich s highly polarized NLSy1s. The following properties are tabulated: optical monochromatic luminosity log νl V in Watts, assuming H 0 = 75kms 1 Mpc 1 and q 0 =0.5; redshift z; p%; FWHM(Hβ b )inkms 1, corrected for instrumental resolution; the intensity ratio [O III]/Hβ b ; Fe II/Hβ b ; rest frame equivalent width EW(Fe II) in Å; the total line flux Hα/Hβ ratio; α opt between 4400Å and 7000Å; α x, the X-ray spectral index with N H unconstrained; N H, the intrinsic soft-x-ray absorption column density in units of cm 2 ; N H,opt is the intrinsic H I column density calculated from E B V (units as for N H). Property CBS 126 IRAS 1239 IRAS 1334 median 90% range Mkn 766 Mkn 1239 IRAS 1509 logνl V (95) z (95) % Pol 1.26± ± ± (43) < 1% 2.34± ± ±0.03 FWHM Hβ b 2850± ± ± (91) ± ± ±200 [O III]/Hβ b 0.29± ± ± (87) ± ± ±0.06 FeII/Hβ b 1.2± ± ± (87) ± ± ±0.1 EW FeII 120±10 175±25 465± (89) ±2 210±10 160±10 Hα/Hβ 3.7± ± ± (70) ± ± ±0.5 α opt 1.0± ± ± (88) ± ± ±0.1 α X ± ± ± (95) ± ± ±1.0 N H ± ± ± (95) ± ± ±4.6 N H,opt X-ray data sources: 1 RASS, Grupe et al. (1998). α x assumes N H = N H,gal 2 Pointed observation. Data retrieved from the MPE ROSAT archive 3 Very High State from Molendi & Maccacaro (1994) Fig. 4. Wavelength dependence of polarization in IRAS F (circles) and IRAS (triangles). Wavelengths (Å) are in the observed frame. fixed to N H,gal (Rush & Malkan 1996). IRAS has shown a probable variation from 0.2 cts s 1 to 0.3 cts s 1 over 3 days of ROSAT pointed observations, compared with 0.37 cts s 1 in the RASS 2.5 years earlier. 4. Discussion 4.1. Statistics of soft-x-ray NLSy1s We found 40 of 43 soft-x-ray AGN to have p <1%, consistent with no intrinsic polarization. Most objects therefore have a direct, unobscured line-of-sight to the nucleus. Some arguments in favor of this can be summarized: Lack of cold X-ray absorption along the line-of-sight. The distribution of N H is basically consistent with the errors in N H derived from spectral fits with N H unconstrained (Grupe et al. 1998). Rapid X-ray variability is common in this sample, suggesting an emission source less than a few light-days in size (Grupe et al. 1998; see also Boller et al. 1993, 1996, 1997). This argues for a direct view of a tiny powerful nuclear source of X-rays and against an extended source such as is expected for a scattering origin. The flat optical spectra and small Balmer line ratios (Hα/Hβ) imply no significant dusty gas. The lack of significant optical linear polarization suggests a dustless line-of-sight. These arguments for a direct unobscured view support the hypothesis, based on spectral energy distributions, that we view the inner accretion disk. Polarization and reddening are usually associated with cold, dusty gas. What makes our soft-x-ray AGN sample interesting is that these sources have steep X-ray spectra and show no soft X-ray absorption by cold, neutral matter, at least in the RASS spectra (Grupe et al. 1998). Thus, it is a surprise that we find any polarized sources at all but we find three two of which clearly show scattering polarization and signs of UV-optical obscuration. These are not the ones with the narrowest BLR emission lines, or steepest X-ray spectra. Both IRAS F and IRAS have FWHM(Hβ b ) close to the magical NLSy1 borderline of 2000 km s 1.

9 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN The polarized AGN UV-optical reddening Assuming that dust is responsible for the continuum and emission-line reddening in the highly polarized AGN, where is this dust located? The two measures of optical absorption α opt and Hα/Hβ are well correlated (Fig. 2). We show on this figure a line representing equal emission line and continuum reddening with respect to typical, essentially unreddened, values for the X-ray-selected sample as a whole. While we have used total Hα and Hβ line strengths to determine Hα/Hβ in this figure, we note that the four most reddened objects have strong narrow line emission. In de-blending the Hα complex, we derived ratios [N II] λλ6548,6584/[o III] λ5007 of While uncertain, comparison of these with values in Koski (1978) and Cohen (1983) suggests little reddening of this NLR emission. Less reddening of NLR Hα and Hβ would imply even larger Hα b /Hβ b ratios for the broad lines than are shown in the figure. Therefore, while the continuum and emission lines appear about equally reddened, it seems likely that the BLR is actually more reddened than the continuum. If more extensive investigation shows this to be true, then the dust may be more closely associated with the BLR than with a region that obscures both the BLR and the continuum source. A consistency check of BLR reddening is to use the line and continuum reddening determined from the relation shown in Fig. 2 to correct the Hβ b intensity, and thus the ratio [O III]/Hβ b. If the BLR suffers reddening similar to the continuum, EW(Fe II) may not be much affected by reddening. This correction brings all the highly polarized sources into line with the low polarization AGN in Fig. 3, consistent with the [O III] λ5007 being relatively unreddened, and arising in a more spatially extended region such as an ionization cone. The BLR reddening has another interesting implication. The highly polarized AGN cover a wide range of [O III]/Hβ b and EW(Fe II) and, after applying this correction for absorption of Hβ b they follow the same well known inverse ( First Principal Component ) relation between [O III]/Hβ b and EW(Fe II) shown by other AGN (Boroson & Green 1992; Grupe et al. 1998, in preparation). We note that, for AGN in general, nuclear reddening must contribute to the scatter in the strong [O III]/Hβ b EW(Fe II) relation and must partially obscure it in samples unbiased with respect to reddening. Reddening introduces important biases when QSOs are selected by optical-uv brightness and color. It is no accident that three IRAS-discovered QSOs are among our sample of highly polarized, reddened AGN. When corrected for reddening IRAS and IRAS have log νl ν 38.0 (watt). Similarly log νl ν for Mkn 1239 and IRAS becomes 37.2 and IRAS , IRAS , and IRAS are then among the most intrinsically luminous AGN at low redshift Warm absorbers Another way to investigate the dusty regions is to compare the neutral-hydrogen column densities inferred from the intrinsic optical absorption ( N H,opt ) and from the soft-x-ray spectral fitting ( N H ). We derived E B V from Fig. 2 to calculate N H,opt k E B V cm 2 (k = 49, Diplas & Savage 1994). The uncertainties could be cm 2, judging by the scatter in Fig. 2. These N H,opt values are given in Table 4. It can be seen that the optical reddening significantly overpredicts N H in all cases except perhaps IRAS , where the very high soft X-ray absorption may be consistent with the optical reddening. If we had used k = 53(Predehl & Schmitt 1995; Predehl & Klose 1996) or k = 66(Gorenstein 1975) the discrepancy would have been even greater. It is also probable that less-reddened, scattered light contributes to the optical spectra (Sect ). In this case the true reddening along a direct path to the nucleus, and hence N H,opt, would be even greater. The relatively low soft X-ray absorption could be because the X-rays arise in a region more spatially extended than the UV-optical. However, the large amplitude and rapid variability (Sect. 3.3) provides evidence that all or most of the X-ray emission arises within regions light-days to light-months in size. This, together with the high brightness in soft X-rays, argues that we have a direct view to the center in X-rays rather than a view via a parsec-scale scattering region. The relatively low soft X-ray absorption could be because the dust-to-gas ratio is high, or, more likely in this nuclear environment, the dusty gas is at least partially ionized and therefore has greater transparency to soft X-ray photons. The dusty gas is probably related to the warm absorbers detected in the O VII and O VIII K-shell edges near kev. This ionized gas has now been detected in IRAS (Sect ), and Mkn 766 (Leighly et al. 1996). Warm absorbers provide a natural tie-in for the UVoptical reddening, X-ray variability, and bright, unabsorbed soft X-ray spectrum. In fact, for IRAS F , neutral hydrogen absorption plus a single power-law does not provide a good fit to the X-ray data (Grupe et al. 1998). In the residuals we find a depression near 0.8 kev an indication of O VII and O VIII absorption edges. The X-ray spectrum and variability will be discussed in more detail together with the results of our spectropolarimetry for this AGN (Wills et al. 1997, in preparation). Recently, Leighly et al. (1997) show that AGN with warm absorber features in their X-ray spectra tend to be significantly polarized. Related to this, Reynolds (1997) also report that Seyfert 1s with UV absorption also showed X-ray warm absorption. Walter & Fink (1993) should be given credit for pointing out that some reddened, highly polarized NLSy1s showed anomalously low νf ν (1375Å) on a plot of Γ vs. νf ν (1375Å)/νF ν (2 kev) (their Fig. 8). The low UV flux (by 30) could be explained on the basis of their observed optical reddening, but they found little or no soft X-ray absorption. All except one are NLSy1s, and now known to show warm X-ray absorption. They are IRAS , Mkn 766, NGC 4051 (Table 1,

10 836 D. Grupe et al.: Scattering and absorption in soft X-ray selected AGN Komossa & Fink 1997), MCG (FWHM(Hβ b ) 1700 km s 1, Pineda et al. 1980, Reynolds et al. 1997), Akn 564 (Table 3). The other NLSy1s discussed here IRAS , IRAS , and Mkn 1239 follow the same trend. When corrected for a reddened continuum all these NLSy1s show the strong increase of Γ with νf ν (1375Å)/νF ν (2 kev) that led Walter & Fink to argue for a UV to soft X-ray bump in Seyfert 1 nuclei Scattered light models We have some evidence that the scattering geometry for NLSy1 may be axisymmetric, and this is certainly true for many other AGN. The polarization (E-vector) is parallel to the major axis of the galaxy in the case of IRAS , and for Mkn 766 and Mkn 1126 the polarization is perpendicular to the elongated radio structure thought to define the jet direction or angular momentum of the central engine (Ulvestad et al. 1995). Mkn 766 shows an ionization cone in the jet direction (Wilson 1997). If we can assume axisymmetry we can relate scattered and direct views to the geometry of the central engine for example, accretion disk geometry and kinematics. IRAS is the best studied of the highly polarized AGN (Wills et al. 1992b; Brandt et al. 1996, 1997; Brinkmann et al. 1996; Hines 1994). In this AGN the contribution of hostgalaxy starlight is <15%, so the wavelength dependence of polarization (Fig. 4) is explained by dilution of polarization by reddened direct AGN light. Therefore we see the AGN from two directions a direct view, and a view from the vantage point of the scatterers. The polarization E-vector is parallel to the host galaxy s major axis, suggesting an axisymmetric geometry with obscuring dust in the plane of the galaxy, probably a dusty torus. Hβ b and Hα may be slightly narrower in scattered than in total light (Hines 1994). X-ray emission in the ROSAT band is strong and rapidly variable, and therefore not seen reflected from a large (parsec-scale) scattering region, but is instead nuclear light seen directly. The steep X-ray spectrum, lack of neutral absorption in X-rays, but strong UV-optical reddening, led Brandt et al. (1996, 1997) to suggest that ionized, dusty gas absorbs the nuclear light, thus leading to the first clear detection of warm absorbers in a luminous QSO. Both the fact that the central AGN is still optically visible, and that the absorption is warm, suggest a line-of-sight to the center that grazes the dusty torus: the absorption optical depths are significant but not extremely high; grains are not likely to have formed in the ionized gas, but pre-existing grains could have evaporated from the inner torus on exposure to the nuclear continuum. A similar scattered-light dusty torus model is a plausible explanation of our second highly polarized object, IRAS F In this case direct sky survey images show that unpolarized host-galaxy starlight is likely to contribute to the wavelength dependence of p% (Fig. 4). Warm absorbers provide a natural explanation for the UV-optical reddening, X-ray variability, and bright, essentially unabsorbed soft X-ray spectrum. Goodrich s (1989) spectropolarimetry suggests a similar scattering geometry for the three highly polarized AGN in his sample of 17 optically selected NLSy1s (Tables 4 and 3), and our investigation of their UV-optical reddening and X-ray spectra suggest both cold and warm absorbers along a direct path to the nucleus. The prototypical NLSy1, I Zw 1, is an interesting example. Because of its extreme NLSy1 properties, NLSy1 are sometimes called I Zw 1 AGN. It was too weak to be included in our X-ray sample because of anomalously high Galactic HI absorption and variability. Its HR1 was during the RASS, but 0.25 during later pointed observations (Boller et al. 1996). It shows significant scattering polarization, with polarized continuum and broad lines (1.7%, Smith et al. 1997) variable in position angle and wavelength-dependence on several-year time scales maybe shorter and possibly significant optical reddening, E(B-V) 0.2. Its steep soft-x-ray spectrum (α x 2.0) is rapidly variable, with very small intrinsic cold absorption (1.5 ± cm 2 ). If reddened, this is likely to be another example of a warm absorber and on axisymmetric Unified Schemes, would therefore be a candidate for a torus-grazing line-of-sight a potentially powerful illustration that steep X- ray spectra need not be observed from the same direction as the narrow Hβ b (and other I Zw 1 properties that go along with these properties) The nature of NLSy1s In principle, comparison of scattered and direct views allows a test for the predicted axisymmetry in kinematics, UV- and X-ray emission (Sect. 1). The important conclusion from Goodrich s (1989) spectropolarimetry is that the narrow Hβ in NLSy1 is more highly polarized than [O III], whose polarization is often unmeasureably small; Hβ b is narrow not because it arises in the NLR; it is produced in separate, higher density gas. Assume that polarized Hβ b represents a scattered light, polar view. Broader polarized Hβ b than that seen in total light could then be interpreted as evidence for scattering by electrons in hot ionized nuclear gas (Goodrich 1989), or for a dust-scattered view of a high-velocity inner BLR and central continuum source that are obscured in a direct high-inclination, view. Narrower polarized Hβ b could be interpreted as arising from a polar view of a flattened BLR. In general, in the 5 cases where a comparison is possible, Goodrich does not find significantly broader Hβ b in polarized light, and therefore favors scattering by cool dusty gas. This assumes that a predominantly direct view of the center is seen in total light; without knowing p% for the scattered light alone, we cannot be sure. Also from the similar widths in polarized and total light, we could argue that a higher-velocity, inner BLR is not hidden by the thicker equatorial regions of a dusty torus, and that there is no evidence for an anisotropic velocity field. Mkn 1239 is an interesting exception. Goodrich finds broad, red wings as an additional feature of the Hβ and Hα broad lines